Letter-search (LS) within a prime often eliminates semantic priming. In 2 lexical decision experiments, the authors found that priming from LS primes occurred for low-frequency (LF) but not high-frequency (HF) targets whether the target's word frequency was manipulated between or within participants and whether the prime-target pairs were associated symmetrically or forward asymmetrically. For the LF targets, LS priming was (a) equivalent for forward asymmetric and symmetric pairs and (b) equal to silent-read (SR) priming for forward asymmetric pairs but less than SR priming for symmetric pairs. The typical finding of greater SR priming for response times for LF than for HF targets occurred for symmetric priming but not for forward asymmetric priming, which showed the interaction for errors. The authors consider their findings' implications for various accounts of how LS affects priming and explain the findings within J. H. Neely and D. E. Keefe's (1989) 3-process model as follows: (a) LS eliminates expectancy and semantic matching but does not reduce semantic activation and (b) expectancy contributes to SR priming for HF targets but not for LF targets, whereas the opposite is so for semantic matching.